Legal claims defining the scope of protection. Each claim is shown in both the original legal language and a plain English translation.
1. A method for real-time processing of IoT data, the method comprising: receiving, by a first physical processor of an edge computing device, a set of data from a first IoT device communicably coupled to the edge computing device, the set of data comprising a set of video data; splitting, by the first physical processor, the set of data into a set of individual data packets; determining, by a second physical processor of the edge computing device, a number of a plurality of instances of the second physical processor for processing the set of individual data packets based on a frame rate of the set of video data and at least one of a number of processor cores available and a bandwidth rate at which the data packets are being received; and processing, by the second physical processor of the edge device, the set of individual data packets by: concurrently applying, by the plurality of instances of the second physical processor of the edge computing device, a learning model to each of a corresponding plurality of data packets from the set of individual data packets; annotating, by a subset of the plurality of instances of the second physical processor, a corresponding subset of the plurality of data packets with a corresponding output from the concurrent application of the learning model; and processing the annotated subset of the plurality of data packets by performing at least one of: object counting, visual click-through analysis, augmented reality facilitation, prediction, and estimation.
This invention relates to real-time processing of IoT data, specifically video data from IoT devices, using edge computing to improve efficiency and reduce latency. The system addresses the challenge of handling high-bandwidth video streams by distributing processing across multiple processor instances at the edge, close to the data source. An edge computing device receives video data from an IoT device and splits it into individual data packets. A second processor determines the optimal number of processing instances based on the video frame rate, available processor cores, and data reception bandwidth. The data packets are then processed concurrently by multiple instances of the second processor, each applying a learning model to its assigned packet. The processed packets are annotated with the model's output, and the annotated data is further analyzed for tasks such as object counting, visual click-through analysis, augmented reality facilitation, prediction, or estimation. This approach ensures low-latency, scalable processing of video data from IoT devices, enabling real-time applications without relying on cloud-based systems.
2. The method of claim 1 , wherein the set of data is split responsive to decoding the set of video data.
This invention relates to video data processing, specifically methods for splitting video data sets during decoding. The problem addressed is the need to efficiently partition video data for parallel processing or storage optimization during decoding, improving performance and resource utilization. The method involves decoding a set of video data and then splitting the decoded data into subsets based on predefined criteria. The splitting process is triggered directly by the decoding operation, ensuring that the data is divided in a way that aligns with the decoded structure. This approach allows for more efficient handling of video data, such as enabling parallel processing of different subsets or optimizing storage by separating data into manageable chunks. The splitting criteria may include temporal segments (e.g., frames or groups of frames), spatial regions (e.g., tiles or blocks), or other logical divisions relevant to the video content. The method ensures that the split subsets maintain coherence and integrity, preserving the necessary relationships between data elements. This is particularly useful in applications requiring real-time processing, such as video streaming, where timely and efficient data handling is critical. By integrating the splitting operation with the decoding process, the invention avoids the need for separate post-processing steps, reducing latency and computational overhead. The method is applicable to various video codecs and formats, making it versatile for different use cases in video compression, transmission, and playback systems.
3. The method of claim 1 , wherein annotating a data packet comprises adding visual effects to an individual data packet.
This invention relates to network data visualization, specifically methods for enhancing the visibility and interpretability of data packets in network traffic analysis. The core problem addressed is the difficulty in identifying and analyzing individual data packets within high-volume network traffic, where traditional methods lack visual distinction and clarity. The method involves annotating data packets by adding visual effects to each packet. These visual effects may include color coding, highlighting, or other graphical modifications to distinguish packets based on their content, source, destination, or other attributes. By visually differentiating packets, analysts can more easily track specific data flows, detect anomalies, or monitor network performance in real time. The annotation process is applied to individual data packets, ensuring that each packet is uniquely marked for visibility. This approach improves the efficiency of network troubleshooting, security monitoring, and performance optimization by providing a clearer representation of network activity. The visual effects can be dynamically adjusted based on user preferences or predefined rules, allowing for customizable and adaptive visualization. This method is particularly useful in environments where network traffic is complex or high-volume, such as in cybersecurity, network diagnostics, or large-scale data transmission scenarios. The visual enhancements help reduce cognitive load on analysts by making critical information more accessible and interpretable.
4. The method of claim 1 , further comprising: responsive to an instance of the second physical processor applying the learning model to an individual data packet, receiving, by the instance, a next data packet of the set of individual data packets for processing.
This invention relates to distributed processing systems where multiple physical processors apply a learning model to individual data packets. The problem addressed is efficiently managing the flow of data packets across distributed processors to ensure continuous and optimized processing without bottlenecks or idle resources. The system includes a first physical processor that distributes a set of individual data packets to multiple instances of a second physical processor. Each instance applies a pre-trained learning model to process an individual data packet. After processing one data packet, an instance automatically receives the next data packet from the set for further processing. This ensures seamless and continuous operation, preventing delays or resource underutilization. The learning model may be a machine learning or deep learning model, and the data packets may contain structured or unstructured data, such as text, images, or sensor readings. The distribution and processing steps are designed to balance workloads across the distributed processors, improving overall system efficiency and throughput. The method may also include error handling, where failed processing attempts trigger re-routing or retry mechanisms to maintain system reliability.
5. The method of claim 1 , further comprising: responsive to annotating each of the plurality of data packets, re-ordering the plurality of data packets to match an order in which the set of individual data packets were processed.
This invention relates to data packet processing in network communication systems, specifically addressing the challenge of maintaining packet order after annotation or modification. In network transmissions, data is often divided into packets, which may be processed or annotated individually. However, this can disrupt the original sequence of packets, leading to misalignment when reassembled at the destination. The invention provides a solution by re-ordering the annotated packets to restore their original sequence, ensuring accurate reconstruction of the transmitted data. The method involves processing a set of individual data packets, annotating each packet with metadata or other modifications, and then re-ordering them to match the sequence in which they were originally processed. This ensures that the annotated packets maintain their correct order, preventing errors in data reconstruction. The re-ordering step is triggered automatically after annotation, streamlining the process and reducing the risk of manual errors. This approach is particularly useful in applications where packet sequence integrity is critical, such as real-time communication, multimedia streaming, or secure data transmission. By preserving the original order, the invention enhances reliability and accuracy in data processing and transmission.
6. The method of claim 5 , further comprising: responsive to re-ordering the plurality of data packets, transmitting the annotated, re-ordered plurality of data packets to a server communicably coupled to the edge device.
This invention relates to data packet processing in network communication systems, specifically addressing the challenge of efficiently managing and transmitting data packets from an edge device to a server. The method involves re-ordering a plurality of data packets based on predefined criteria, such as priority, latency requirements, or network conditions, to optimize transmission efficiency and reduce latency. After re-ordering, the data packets are annotated with metadata that may include timestamps, priority indicators, or routing information to facilitate proper handling by the receiving server. The annotated, re-ordered data packets are then transmitted to a server communicably coupled to the edge device, ensuring that the data is processed in an optimized sequence. This approach improves network performance by minimizing delays and ensuring that critical data is prioritized during transmission. The method may also include additional steps such as error checking, packet validation, or dynamic adjustment of re-ordering criteria based on real-time network conditions to further enhance reliability and efficiency. The overall system ensures that data is transmitted in a structured and optimized manner, reducing bottlenecks and improving communication between edge devices and central servers.
7. The method of claim 1 , further comprising: determining the learning model to apply based on a set of characteristics of the set of data.
This invention relates to machine learning systems that adaptively select learning models based on data characteristics. The problem addressed is the inefficiency of applying a fixed learning model to diverse datasets, which can lead to suboptimal performance. The solution involves dynamically determining the most suitable learning model for a given dataset by analyzing its characteristics, such as statistical properties, feature distributions, or structural patterns. This adaptive selection ensures that the chosen model aligns with the data's inherent properties, improving accuracy and efficiency. The method may involve preprocessing the data to extract relevant characteristics, comparing these characteristics against predefined criteria or model performance metrics, and selecting the model that best matches the data's requirements. The system may also include training multiple candidate models and evaluating their performance on the dataset to make an informed selection. By tailoring the learning model to the data, this approach enhances predictive accuracy and computational efficiency, particularly in scenarios where datasets vary significantly in structure or complexity. The invention is applicable in fields such as predictive analytics, natural language processing, and computer vision, where model performance is highly dependent on data characteristics.
8. A non-transitory machine-readable storage medium comprising instructions executable by a physical processor of an edge device for real-time processing of IoT data, the machine-readable storage medium comprising: instructions to cause a first physical processor to receive a set of data from a first IoT device communicably coupled to the edge device, the set of data comprising a set of video data; instructions to cause the first physical processor to split the set of data into a set of individual data packets; instructions to cause a second physical processor to determine a number of a plurality of instances of the second physical processor for processing the set of individual data packets based on a frame rate of the set of video data and at least one of a number of processor cores available and a bandwidth rate at which the data packets are being received; instructions to cause the second physical processor to process the set of individual data packets by: concurrently applying, by the plurality of instances of the second physical processor, a learning model to each of a corresponding plurality of data packets from the set of individual data packets; annotating, by a subset of the plurality of instances of the second physical processor, a corresponding subset of the plurality of data packets with a corresponding output from the concurrent application of the learning model; and processing the annotated subset of the plurality of data packets by performing at least one of: object counting, visual click-through analysis, augmented reality facilitation, prediction, and estimation; and instructions to transmit the processed set of individual data packets to a server communicably coupled to the edge device.
This invention relates to real-time processing of IoT data, specifically video data, at the edge of a network to reduce latency and bandwidth usage. The system involves an edge device with multiple processors that receives video data from an IoT device. The data is split into individual packets for parallel processing. A second processor determines the optimal number of processing instances based on the video frame rate, available processor cores, and incoming data bandwidth. These instances concurrently apply a learning model to different data packets. A subset of the processors annotates the processed packets with model outputs, which are then used for tasks like object counting, visual click-through analysis, augmented reality, prediction, or estimation. The processed data is transmitted to a server. This approach enables efficient, low-latency video analysis at the edge, reducing the need for cloud-based processing and improving real-time decision-making.
9. The non-transitory machine-readable storage medium of claim 8 , wherein the instructions to cause the first processor to split the set of data comprises instructions to split the set of data responsive to decoding the received set of data, and wherein the storage medium further comprises: instructions to cause the processed set of individual data packets to be re-encoded.
This invention relates to data processing systems, specifically methods for handling and encoding data packets. The technology addresses the challenge of efficiently managing and processing large sets of data by splitting them into smaller, individual data packets for further processing. The system involves a non-transitory machine-readable storage medium containing instructions that, when executed by a processor, perform specific operations. The instructions cause a first processor to split a received set of data into individual data packets after decoding the data. The split data packets are then processed, and the processed packets are re-encoded before further use or transmission. This approach improves data handling efficiency by breaking down large datasets into manageable units, facilitating parallel processing or optimized storage. The re-encoding step ensures the processed data maintains integrity and compatibility with subsequent systems or applications. The invention is particularly useful in environments where data must be processed in segments, such as distributed computing systems, real-time data streams, or large-scale data storage solutions. The method ensures that data is accurately split, processed, and re-encoded, maintaining reliability and performance in data-intensive operations.
10. The non-transitory machine-readable storage medium of claim 8 , wherein the instructions to cause the second physical processor to annotate a data packet comprise: instructions to cause the second physical processor to add visual effects to an individual data packet.
A system for processing data packets in a network environment addresses the challenge of enhancing data packet visibility and interpretability during transmission. The system includes a first physical processor that receives a data packet from a source device and a second physical processor that annotates the data packet before forwarding it to a destination device. The annotation process involves adding visual effects to the data packet, such as color coding, highlighting, or other graphical modifications, to improve readability and analysis. The visual effects are applied to individual data packets, allowing for distinct identification and tracking. The system may also include a memory storing instructions for executing these processes, ensuring efficient and accurate data packet handling. This approach enhances network monitoring, debugging, and security by providing clearer insights into data flow and packet characteristics. The visual annotations help operators quickly identify packet types, priorities, or anomalies without requiring extensive decoding or analysis. The system is particularly useful in high-traffic networks where rapid packet identification is critical for performance optimization and troubleshooting.
11. The non-transitory machine-readable storage medium of claim 8 , further comprising: instructions to cause the second physical processor to, responsive to an instance of the second physical processor applying the learning model to an individual data packet, receive, by the instance, a next data packet of the set of individual data packets for processing.
A system for distributed processing of data packets using a learning model involves multiple physical processors, each executing an instance of the learning model to analyze individual data packets. The system addresses the challenge of efficiently processing large volumes of data by distributing the workload across multiple processors, improving throughput and reducing latency. Each processor instance applies the learning model to a data packet and then automatically receives the next data packet in the sequence for further processing. This continuous flow ensures minimal idle time between processing tasks, optimizing resource utilization. The learning model is trained to perform specific tasks such as classification, prediction, or anomaly detection on the data packets. The system may also include mechanisms to manage data distribution, load balancing, and synchronization between processors to maintain consistency and accuracy in the processing results. The approach is particularly useful in high-performance computing environments where real-time or near-real-time processing of data streams is required, such as in network monitoring, financial transactions, or industrial automation. The use of a non-transitory machine-readable storage medium ensures that the instructions for executing the learning model and managing the data flow are persistently stored and accessible to the processors.
12. The non-transitory machine-readable storage medium of claim 8 , further comprising: instructions to cause the second physical processor to, responsive to annotating each of the plurality of data packets, re-order the plurality of data packets to match an order in which the set of individual data packets were processed.
This invention relates to data processing systems that handle multiple data packets, particularly in scenarios where packets are processed out of their original order. The problem addressed is maintaining the correct sequence of data packets after processing, which is critical for applications requiring ordered data, such as real-time communication, financial transactions, or network protocols. The system involves a non-transitory machine-readable storage medium containing instructions for a second physical processor. After processing a set of individual data packets, the system annotates each packet with metadata indicating its position or sequence. The processor then re-orders the annotated packets to restore their original sequence, ensuring that the output matches the order in which the packets were initially processed. This re-ordering step is essential when packets are processed in parallel or out of sequence, as it corrects any disruptions caused by asynchronous processing. The invention ensures data integrity and consistency by preserving the original order of packets, which is particularly useful in systems where timing and sequence are critical. The annotations may include timestamps, sequence numbers, or other identifiers that facilitate accurate re-ordering. The system may also include additional instructions for error detection or correction to further enhance reliability. This approach improves performance in distributed systems, high-speed networks, and other environments where maintaining packet order is necessary.
13. A system for real-time processing of IoT data, the system comprising: a first physical processor of an edge device that implements machine readable instructions that cause the system to: receive a set of video data from a first IoT device communicably coupled to the edge device; split the set of data into a set of individual frames including data packets; transmit the set of individual frames to a second physical processor of the edge device, wherein the second physical processor is to implement machine readable instructions that cause the system to: determine a number of a plurality of instances of the second physical processor for processing the set of individual frames based on a frame rate of the set of video data and at least one of a number of processor cores available and a bandwidth rate at which the data packets are being received; and process the set of individual frames by: concurrently applying, by the plurality of instances of the second physical processor, a learning model to each of a corresponding plurality of frames from the set of individual frames; annotating, by each of the plurality of instances of the second physical processor, each of the plurality of frames with a corresponding output from the concurrent application of the learning model; and processing the annotated plurality of frames by performing at least one of: object counting, visual click-through analysis, augmented reality facilitation, prediction, and estimation.
The system is designed for real-time processing of IoT data, specifically video data from IoT devices, using edge computing to reduce latency and improve efficiency. The problem addressed is the need for scalable, low-latency processing of high-frame-rate video streams from IoT devices, which traditional cloud-based systems cannot handle effectively due to bandwidth and processing delays. The system includes an edge device with at least two physical processors. The first processor receives video data from an IoT device and splits it into individual frames, each containing data packets. These frames are then transmitted to the second processor, which dynamically determines the number of processing instances required based on the video frame rate, available processor cores, and data packet bandwidth. The second processor distributes the frames across multiple instances, each applying a machine learning model to analyze the frames concurrently. Each instance annotates its assigned frames with the model's output, such as object detection or classification results. The annotated frames are then further processed for tasks like object counting, visual click-through analysis, augmented reality, prediction, or estimation. This distributed approach ensures real-time performance while optimizing resource usage.
14. The system of claim 13 , wherein the first physical processor implements machine readable instructions to cause the system to: split the set of data into the set of individual frames responsive to decoding the received set of data; and re-encode the processed set of data.
This invention relates to a data processing system designed to handle and transform sets of data, particularly in applications involving video or multimedia streams. The system addresses the challenge of efficiently processing large data sets by breaking them into smaller, manageable units for further manipulation. The system includes a first physical processor that executes machine-readable instructions to perform specific operations. Initially, the system receives a set of data, which may be encoded in a compressed or proprietary format. The first physical processor decodes this received data to extract its raw or uncompressed form. Once decoded, the system splits the data into a set of individual frames, which are discrete units of the original data set. These frames are then processed according to predefined criteria, such as filtering, compression, or transformation. After processing, the system re-encodes the modified set of data into a desired output format, ensuring compatibility with downstream applications or storage systems. The re-encoding step may involve applying lossless or lossy compression techniques, depending on the requirements of the application. The system is designed to handle these operations in real-time or near real-time, making it suitable for applications like video streaming, surveillance, or multimedia editing. The invention improves upon existing systems by integrating decoding, frame-based processing, and re-encoding into a unified workflow, reducing latency and improving efficiency in data transformation tasks.
15. The system of claim 13 , wherein the second physical processor implements machine readable instructions to cause the system to: annotate an individual frame by adding visual effects to the individual frame.
This invention relates to a computer system for processing video frames, specifically enhancing individual frames with visual effects. The system includes at least two physical processors working in tandem. The first processor is responsible for capturing or receiving video frames, while the second processor applies machine-readable instructions to modify these frames. The second processor annotates each frame by adding visual effects, such as overlays, filters, or other graphical enhancements. These effects may include text, graphics, or dynamic elements that alter the appearance of the frame. The system ensures that the visual effects are applied in real-time or near-real-time, maintaining synchronization with the original video stream. The invention addresses the need for efficient, automated frame annotation in applications like video editing, surveillance, or augmented reality, where real-time processing and visual enhancement are critical. The use of separate processors optimizes performance by distributing computational tasks, allowing for faster processing and reduced latency. The system may also include additional components, such as memory storage for temporary or permanent storage of processed frames, and input/output interfaces for receiving and transmitting video data. The invention improves upon existing methods by providing a scalable, high-performance solution for real-time frame annotation.
16. The system of claim 13 , wherein the second physical processor implements machine readable instructions to cause the system to: responsive to an instance of the second physical processor applying the learning model to an individual frame, receive, by the instance, a next frame of the set of frames for processing.
This invention relates to a distributed processing system for analyzing video frames using machine learning models. The system addresses the challenge of efficiently processing large volumes of video data by distributing the workload across multiple physical processors. Each processor applies a learning model to individual frames of a video sequence, enabling parallel processing to improve throughput and reduce latency. The system includes a first physical processor that distributes frames from a video stream to a second physical processor for analysis. The second processor applies a machine learning model to each frame, generating results such as object detection, classification, or other inferences. After processing a frame, the second processor automatically receives the next frame in the sequence for continued analysis, ensuring continuous and uninterrupted workflow. The distributed architecture allows for scalable and high-performance video processing, suitable for applications like real-time surveillance, autonomous systems, or video analytics. The system may also include additional processors or components to manage data flow, optimize resource allocation, or enhance processing efficiency. The invention focuses on seamless frame-by-frame processing without manual intervention, leveraging parallel computing to handle high-resolution or high-frame-rate video streams effectively.
17. The system of claim 13 , wherein the second physical processor implements machine readable instructions to cause the system to: responsive to annotating each of the plurality of frames, re-order the plurality of frames to match an order in which the set of individual frames were processed.
This invention relates to a video processing system that annotates and reorders video frames. The system addresses the challenge of maintaining temporal consistency in video analysis when frames are processed out of their original sequence. The system includes a first physical processor that processes a set of individual video frames in parallel, annotating each frame with metadata such as object detection, tracking, or classification results. A second physical processor then reorders the annotated frames to restore their original chronological sequence, ensuring that the output video maintains temporal coherence. This reordering step is crucial for applications like surveillance, autonomous driving, or video analytics, where the sequence of events must be preserved. The system improves efficiency by parallelizing frame processing while ensuring the final output retains the correct temporal order. The invention also includes mechanisms to handle frame dependencies, such as tracking objects across frames, by aligning annotations with the original frame sequence. This approach optimizes computational resources while maintaining accurate temporal relationships in the processed video.
Unknown
August 27, 2019
Browse 5M+ US patents with plain-English claim translations and AI-generated analysis.